The development of a key poses selector for use in 3D games

However, the drawback of pose-base recognition systems is that users may simply adopt a single pose without fully completing the movement and that pose will trigger the event. J.Y. Wang 2009 and C.T. Hsieh et al. 2012 applied three-dimensional coordinates to classify under Support Vector Machine (SVM). This approach only failed when two animations were similar, for example, “walk” and “run”. In order to improve the drawback of a pose-based system, T.Y. Lin et al. 2013 applied four poses in each movement. This approach improved the movement detection.